When I export a Catalyst Python V2 script from Paraview 5.9.1 and try to run it as an in-situ pipeline in a ParaView Catalyst instrumented code, the corresponding Input Data Description always seems to request all fields rather than just the ones needed by the pipeline which can be a little costly in a simulation and makes code that checks what fields are needed (e.g. this line) somewhat redundant (at least for these scripts)?
I can see from a comment in the vtkCPPythonScriptV2Helper class that this is because these new scripts do not implement data descriptions like the old scripts. Is there a plan to implement these soon so that my simulations don’t pass fields unnecessarily to Catalyst?
I vaguely remember this issue being fixed after the 5.9.1 release. I believe there is a restriction though that for the first call that all fields and meshes are requested for the first time step.
Any chance you can try this out with the 5.10 release to see if this is fixed and working properly?
Whilst browsing the GitLab repository, I came across this merge request which seems to be along the lines of what you were saying but this seems to be present in v5.9.1.
In any case, I will try building v5.10.0 and report back.
Sorry for the slow reply. I just tried out v5.10.0 and still have the same problem so I don’t think it’s been fixed. Any idea if a solution is in the works?